Engineers at the Massachusetts Institute of Technology have developed an artificially intelligent "food computer" that automatically optimizes conditions, such as humidity and light, to produce the desired qualities in plants. Unlike other similar developments, the system is open source, and engineers have posted instructions on how to assemble and configure it on their own. A detailed description and documentation are available on the MIT Media Lab website, and Fast Company also writes about it.
In order to increase the efficiency of growing plants by reducing the space required for this, various concepts are being developed, such as vertical farms and growing in a controlled environment. Unlike conventional agriculture, the plants are not illuminated by sunlight, but by LEDs. Some are trying to adapt them for mass deployment and self-assembly.
MIT engineers presented a prototype of their system back in 2015, and called it OpenAg. The prototype consisted of a metal container, inside which were located sensors, LEDs, pumps for supplying water and fertilizers, and other components that allowed changing the climate inside the container. They also presented a more massive version, placed in a standard cargo container, and consisting of several cells, in each of which separate conditions were maintained. Initially, scientists independently analyzed the data and selected the best, in their opinion, conditions.
In mid-2016, the research team began partnering with an artificial intelligence company. The goal of the partnership was to speed up and automate the selection of the necessary conditions. The researchers chose basil as a model plant. The system continuously analyzed the condition of the plants from the parallel working cells and made adjustments to the conditions for growing the next crop.
Array of several isolated cells with different conditions / Open Agriculture Initiative, MIT Media Lab (openag.mit.edu | CC-BY-SA 4.0)
The researchers were able to optimize the algorithm in order to increase the synthesis of specific molecules responsible for taste in the plant several times. In the future, they plan to teach the system how to optimize conditions for many different characteristics such as taste, size or cost. The researchers especially note that, unlike other well-known developments in this area, their system is completely open and available for self-construction or modification.